What the log says so far. Every number here is counted from the log when the site builds, so one more morning moves them all.
Graded by Eye
Two Clouds Decide It
A great sunrise needs cloud overhead and a clear path toward the sun. With a full deck overhead and the path clear, 15 of 19 mornings were great. With nothing overhead, 2 of 32. With cloud in the way toward the sunrise, 13 of 21 were flat.
I graded the sunrise photograph of 117 mornings by eye: 27 great, 55 with some color, 35 flat. Each morning is then sorted by the weather archive's clouds 10 minutes before its sunrise, into the first sky that fits: low cloud overhead, then low and mid cloud 100 to 300 km toward the sunrise, then how much high or mid cloud was overhead. Ranked by the cloud toward the sunrise, a morning with color came out ahead of a flat one 7 times in 10. Ranked by the deck overhead, a great morning came out ahead of the rest 8 times in 10.
Why cloud 100 km away matters, the photographs under each sky and a simulator to try it are on What Makes a Good Sunrise?
Color Index
Against the Cloud Layers
The photographs held less color under low and mid cloud and more under high cloud. With low cloud over 50 percent of the sky the median color index was 30.5, against 35.6 without. Mid cloud: 26.6 against 35.0. High cloud: 36.4 against 33.4.
Over 50 percent: 30.5 across 19 mornings. At or under it: 35.6 across 111.
Over 50 percent: 26.6 across 10 mornings. At or under it: 35.0 across 120.
Over 50 percent: 36.4 across 51 mornings. At or under it: 33.4 across 79.
One dot per photographed morning, placed by its color index and by the share of the sky that layer covered at sunrise, from the weather archive. The fill is the morning's low and middle cloud together, orange when it was clear and gray when it was overcast. The dashed line is where the mornings are split, the green bars are the medians either side of it, and each dot links to its hike. The index ran from 0.0 to 73.8 across these mornings, so differences this size are leanings, not rules.
Color Index
By Month
July had the highest median color index, 38.4 across 9 mornings, and December the lowest, 22.9 across 5.
Every photographed morning in its calendar month, pooled across years, with the count under each month. A month with at least 5 mornings gets a bar from its lowest to its highest and a tick at the median; the rest show their dots and nothing more.
12 months have at least 5 photographed mornings. Median color index, highest first: July at 38.4 (9 mornings), June at 38.3 (14), April at 37.9 (10), May at 37.3 (11), September at 35.2 (14), October at 34.8 (14), August at 34.6 (15), November at 33.3 (11), February at 32.5 (9), January at 29.9 (10), March at 29.4 (8), December at 22.9 (5). The widest spread among them is February, 0.0 to 64.9.
Sunrise
Through the Year
115 of 172 mornings were clear at sunrise, 67 percent. Fall was the clearest season at 76 percent and winter the cloudiest at 50. Sunrise came as early as 6:25 AM (June 13) and as late as 7:57 AM (November 2).
The curve is the computed sunrise for the Cherokee Trail, Stone Mountain sunrise spot in local clock time, the trail with the most logged hikes. Each dot is a logged morning at its own measured sunrise, on every trail: orange when the low and middle cloud was clear, gray when it was overcast, hollow when no weather record exists. High cloud does not count against a morning.
Clear means low and middle cloud together at or under 30 percent, across 172 mornings with a weather record. By season: winter 50 percent of 34 mornings, spring 66 percent of 44, summer 71 percent of 49, fall 76 percent of 45. Among the photographed mornings, the median color index was 36.0 on the 96 clear ones and 31.1 on the 34 others.
The earliest logged sunrise was 6:25 AM on June 13, 2026, the latest 7:57 AM on November 2, 2024, across 172 mornings. On the Cherokee Trail, Stone Mountain, the sun came up anywhere from 61° (June 22) to 118° (December 21) along the horizon. The sunrise sweep marks every morning on the view from the Boatyard dock, with the spread between the computed bearing and the sun in the frame stated as spread.
Without a Chart
Also Counted
When the color comes. 59% of the 135 sunrise photographs I kept were taken before the sun cleared the horizon, and nine in ten came no earlier than 14 minutes before it. By the minute, the median color index was 34.2 in the ten minutes before sunrise (52 frames) and 35.0 in the ten after (29). The wake-up calculator counts back from that.
Coldest at the start. On 96 of 128 recorded walks, the coldest hour of the walk came within an hour of sunrise, and about two hours after sunrise it was a median 3°F warmer (119 walks). The temperatures are the weather archive's hours for the walk, not a thermometer on the trail. What I wear for it is in Start Cold.
Reading This Page
What Counts as a Morning
Nothing on this page is typed in. Every number is computed from the log when the site builds, the charts are drawn from the same rows, and adding one hike moves every figure. A sentence appears only when each group it describes holds at least 5 mornings, so a chart with nothing said about it has not earned a sentence yet.
The log holds 172 mornings with a cloud reading, 130 photographed mornings (median color index 34.7) and 117 mornings graded by eye. A photographed morning is a hike whose front photograph is a sunrise frame with a color index. The weather belongs to the morning, so a morning with two frames counts once, by the frame that fronts its page, and a photograph from later on the trail is not a sunrise and stays out. Every logged hike with a cloud reading counts toward the sunrise chart and the clear-sky share, photographed or not.
The cloud fractions are the weather archive's low, middle and high layers for the sunrise hour, and cloud cover is the low and middle layers combined, the cloud that hides the sun, as everywhere on this site. High cloud is cirrus, the cloud a sunrise lights up, and it is always shown on its own. The color index is described on the methods page, along with the sources and their limits. Medians throughout, because one morning should not move the typical one, and no median or rank is stated from fewer than 5 mornings. The same rows are in the downloadable log, so any figure here can be recomputed elsewhere.